Re: A Chinese student asks for help on DEAL.II

[Wolfgang Bangerth]

Jian,

On 08/22/2013 10:16 AM, yuj...@ase.buaa.edu.cn wrote:
> Dear Prof. Wolfgang Bangerth,
> I am a post doctor in China. I am writing to you for help on DEAL. II. My
> research area is the discontinuous Galerkin method. I am very interested in
> using DEAL. II for my research. And I plan to devolop my DG code with DEAL.II.
> However, I am also concerned with the following questions
> 1) Is DEAL.II able to handle grids for complex geometries such as DLR-F4 from
> the AIAA drag prediction workshop (DPW) ? Also we generally use softwares like
> Gridgen or ICEM ot generate grids. Can DEAL.II handle grids from these softwares?

We do have classes that read meshes that have previously been generated, see
http://www.dealii.org/developer/doxygen/deal.II/classGridIn.html
If you have meshes in different formats, it is not typically very complicated
to add a new reader to this class.

We have used fairly complex meshes in the past. The limitation is that we can
only deal with quadrilaterals and hexahedra. See also

https://code.google.com/p/dealii/wiki/FrequentlyAskedQuestions#Can_I_use/implement_triangles/tetrahedra_in_deal.II?

https://code.google.com/p/dealii/wiki/FrequentlyAskedQuestions#How_do_I_create_the_mesh_for_my_problem?


> 2) Is DEAL.II able to support moving grids with moving boundraies for
> situations like airfoil dynamic stall? Also if there are multi-body seperation
> problems, one may need to regenerate the grids every time step. Is DEAL.II
> able to handle this situation ?

Like so many other problems, the answer is "generally yes, but it depends on
the details". We do have a moving mesh in the step-18 tutorial program.
Regenerating the grid is possible through external programs, but you have to
pay the price of having to interpolate from one mesh to the next. A simpler
approach is typically to use adaptive mesh refinement because then there is a
clear path to interpolating from the old to the new mesh.


> 3) In the near future, I wish to develop a DG code with DEAL.II for
> engineering applications in the area of complex aircraft aerodynamics using
> RANS, DES and LES. Is this idea feasible? Or has anyone used DEAL.II to do
> similar work?

There is Ralf Hartmann's work on similar problems. His code is based on a
deal.II version from ~10 years ago, and it is not publicly available. You may
want to contact him nevertheless.

It is of course feasible to do this with deal.II but, as before, it is a
question of how much work you are willing to put into your application. We
have all of the tools for which you want to do (handling complex meshes, DG
elements, parallel computing) but you will have to implement the model
equations , the time stepping, the turbulence model, etc.

Best
Wolfgang

--
------------------------------------------------------------------------
Wolfgang Bangerth email: bang...@math.tamu.edu
www: http://www.math.tamu.edu/~bangerth/

[Thomas Wick]

Hi Jian,

concerning your second question:

>> 2) Is DEAL.II able to support moving grids with moving boundraies for
>> situations like airfoil dynamic stall? Also if there are multi-body
>> seperation
>> problems, one may need to regenerate the grids every time step. Is
>> DEAL.II
>> able to handle this situation ?
>
> Like so many other problems, the answer is "generally yes, but it
> depends on
> the details". We do have a moving mesh in the step-18 tutorial program.
> Regenerating the grid is possible through external programs, but you
> have to
> pay the price of having to interpolate from one mesh to the next. A
> simpler
> approach is typically to use adaptive mesh refinement because then
> there is a
> clear path to interpolating from the old to the new mesh.
>

The answer is clearly YES. There are some guys in deal.II (including
myself) who compute
fluid-structure interaction. Here, you have
- a moving interface
- moving boundaries
- different physics in different parts of the domain
- there is also a deal.II-step on this under www.archnumsoft.org with
the specific address

http://archiv.ub.uni-heidelberg.de/ojs/index.php/ans/issue/archive

Of coarse, as Wolfgang is saying, it is costly since you have to
regenerate the mesh or
solve an additional PDE in order to move the mesh.


Best Thomas

--
++-------------------------------------------------------++
Dr. Thomas Wick
Postdoctoral Fellow
The University of Texas at Austin
Institute for Computational Engineering and Sciences (ICES)
201 East 24th Street
ACE 5.332 | Campus Mail C0200
Austin, TX 78712, USA

Email: tw...@ices.utexas.edu
Tel.: +1 (512) 232-7763
www: http://numerik.uni-hd.de/~twick/
++-------------------------------------------------------++
--

--
++-------------------------------------------------------++
Dr. Thomas Wick
Postdoctoral Fellow
The University of Texas at Austin
Institute for Computational Engineering and Sciences (ICES)
201 East 24th Street
ACE 5.332 | Campus Mail C0200
Austin, TX 78712, USA

Email: tw...@ices.utexas.edu
Tel.: +1 (512) 232-7763
www: http://numerik.uni-hd.de/~twick/
++-------------------------------------------------------++
--