> Am 16.07.2019 um 08:55 schrieb lc <[email protected]>:
> 
> Hello Timo,
> 
> thanks for your answer.
> 
> Since, I'm working with Dumux 2.12 I cannot say anything about 1) or 2).
> 
> So, for the moment, I attach al the material necessary to replicate the 
> behaviour that I get on Dumux 2.12, at least to understand if it is a bug, my 
> implementation mistake or should be like this, if case you would have time to 
> take a look at it.
> 
Hi Lorenzo,

please post a minimal working example to reproduce the behaviour. Also, we need 
to know what you are computing, what is the outcome, what is the expected 
outcome and why you think it’s a bug. 

Timo

> In the meanwhile, I'll try to port such  in Dumux 3.0 and try to run the box 
> scheme and the cell-centered to see if it goes better.
> 
> 
> Best regards,
> 
> Lorenzo
> 
> 
>> On 15.07.2019 15:31, Timo Koch wrote:
>> Hi Lorenzo,
>> 
>> I think the status is as follows (Dumux 3.0):
>> 
>> 1) For the box scheme Dirichlet boundaries are always strongly enforced so 
>> nothing special to do here.
>> 
>> 2) For cell-centered schemes there is a new feature called "internal 
>> Dirichlet constraints" that allows you to enforce Dirichlet constraints in 
>> any cell. See 
>> https://git.iws.uni-stuttgart.de/dumux-repositories/dumux/blob/master/test/porousmediumflow/1p/implicit/internaldirichlet/problem.hh
>>  for an example (only available on the master branch).
>> 
>> 3) For the old sequential schemes there is _no_ such feature as far as I 
>> know (unfortunately noone had the time to implement sequential schemes in 
>> terms of the Dumux 3.0 style solvers yet, so actually there is no "new" 
>> sequential schemes yet, but I call it "old" because it uses Dumux 2.0 style 
>> paradigms that have been deprecated). You'd have to implement it yourself 
>> (adapt the matrix assembly, in the simplest case just manipulate the rows of 
>> Dirichlet dofs after everything has been assembled, replace row with zeros 
>> set one on the diagonal and the adapt the rhs).
>> 
>> That being said,
>> 
>> I'm not sure why you need this feature in your case. If the boundary 
>> condition is satisfied in the first cell or on the boundary shouldn't really 
>> matter. It's just that in the latter case you don't see the actual boundary 
>> value in your vtk output.
>> 
>> It might be that there is actually something wrong with the boundaries in 
>> the old sequential schemes. I personally don't know of any bug (and there is 
>> no open issue on that topic), but there is always the possibility. In that 
>> case, I would need a minimal working example how to reproduce an actual 
>> erroneous solution (in the best case open a merge request with the minimal 
>> working example). I'm afraid the old sequential schemes don't get very much 
>> attention right now from the core developers.
>> 
>> Sorry I couldn't help more, open-source research software is always a work 
>> in progress,
>> 
>> Timo
>> 
>> 
>>> On 15.07.19 13:15, lc wrote:
>>> Hello,
>>> 
>>> can you explain how can I strongly impose the Dirichet condition in the 
>>> first cell (or the closest one) on the inlet (left ) boundary in order to 
>>> achieve what you said, please?
>>> 
>>> I need to use 2p, sequential algorithm.
>>> 
>>> 
>>> Kind regards,
>>> 
>>> Lorenzo
>>> 
>>> 
>>>> On 29.03.2019 16:31, Timo Koch wrote:
>>>> I mean you can also just enforce the Dirichlet value strongly in the first 
>>>> cell. Then, of course it will be satisfied "cleanly". It all depends on 
>>>> your discretization scheme too. For example for FEM or Box you might have 
>>>> degrees of freedom directly on the boundary. Then you can also strongly 
>>>> enforce them an they will always be "cleanly" satisfied.
>>> 
> 
> -- 
> ___________________________________________
> *Il tuo 5 diventa 1000*
> Fai crescere la tua università
> Dona il 5 per mille alla Sapienza
> Codice fiscale: *80209930587*
> 
> https://www.uniroma1.it/it/pagina/fai-crescere-la-tua-universita-con-il-cinque-mille
>  
> <https://www.uniroma1.it/it/pagina/fai-crescere-la-tua-universita-con-il-cinque-mille>
> 
> <kernA1.tar.gz>
> _______________________________________________
> Dumux mailing list
> [email protected]
> https://listserv.uni-stuttgart.de/mailman/listinfo/dumux

_______________________________________________
Dumux mailing list
[email protected]
https://listserv.uni-stuttgart.de/mailman/listinfo/dumux

Reply via email to