Skip to first unread message

Frederik S.

unread,
Nov 26, 2015, 7:35:06 AM11/26/15
to deal.II User Group
Dear all,

I have a problem with the meshes I import with the GridIn class:
As basis of my calculations I use .tif-Image stacks of a microstructure. In a first step I get a grid where each voxel is represented by one cell of the mesh (over 10 Million cells), which I then coarse with hanging nodes so that the bulk of structures is coarse but the boundaries between them are still very fine (several 100,000 cells). The documentation of the GridIn class says, that an input mesh "must not contain hanging nodes", because "most mesh description formats do not store neighborship information between cells". Is there any way that I can still use the hanging node input mesh? Which format could store these neighbourship information and in which form should it be stored? I currently store my mesh as an .ucd-file and use deal.ii 8.3

Greetings and many thanks in advance,
Frederik

Krzysztof Bzowski

unread,
Nov 27, 2015, 2:41:44 AM11/27/15
to dea...@googlegroups.com
Little off topic. What kind of software did you use to create mesh based
on microstructure?

Best regards,
K.

Frederik S.

unread,
Nov 27, 2015, 3:15:47 AM11/27/15
to deal.II User Group
Little off topic. What kind of software did you use to create mesh based
on microstructure?

Best regards,
K.

Hello Krzysztof,

I use an in house code written by a software engineer also involved in this project.

Best regards,
Frederik

Krzysztof Bzowski

unread,
Nov 27, 2015, 5:35:25 AM11/27/15
to dea...@googlegroups.com
On 2015-11-27 09:15, Frederik S. wrote:
> Hello Krzysztof,
>
> I use an in house code written by a software engineer also involved in
> this project.
>

If the meshing project is also written in C++ you can directly use
deal.ii data structures (like Triangulation) Consider merging meshing
software with FEM software based on deal.ii. Just start from coarse
mesh, refine global few times until you reach pixel dimension and apply
your coarsening/refining approach. You will avoid additional file
transfers and solve the problem with hanging nodes.

K.

Frederik S.

unread,
Nov 27, 2015, 7:09:23 AM11/27/15
to deal.II User Group
If the meshing project is also written in C++ you can directly use
deal.ii data structures (like Triangulation) Consider merging meshing
software with FEM software based on deal.ii. Just start from coarse
mesh, refine global few times until you reach pixel dimension and apply
your coarsening/refining approach. You will avoid additional file
transfers and solve the problem with hanging nodes.

K.


Thank you for this good suggestion, but the meshing code isn't written in C++, so I think this solution would be relativly complex to implement. 
Isn't there any possibility to also store the needed neighborship information between cells in some mesh-file-format?

F.

Timo Heister

unread,
Nov 28, 2015, 9:57:54 AM11/28/15
to dea...@googlegroups.com
> The documentation of the GridIn class
> says, that an input mesh "must not contain hanging nodes", because "most
> mesh description formats do not store neighborship information between
> cells". Is there any way that I can still use the hanging node input mesh?

No, this is not only a limitation in file formats. We also expect the
coarse mesh inside the Triangulation to be conforming. I don't see an
easy way to do what you are trying to do except figuring out a way to
start with a conforming coarse mesh and then adaptively refining it.

--
Timo Heister
http://www.math.clemson.edu/~heister/

Wolfgang Bangerth

unread,
Nov 28, 2015, 3:10:22 PM11/28/15
to dea...@googlegroups.com
On 11/28/2015 08:57 AM, Timo Heister wrote:
> No, this is not only a limitation in file formats. We also expect the
> coarse mesh inside the Triangulation to be conforming. I don't see an
> easy way to do what you are trying to do except figuring out a way to
> start with a conforming coarse mesh and then adaptively refining it.

Correct. I think you'll have to define your own file format in which you store
both the coarse mesh and how to refine it in order to come up with the mesh
you want. You'll then need a (relatively simple) function in your application
that reads this file and refines the mesh accordingly.

Best
W.

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

Frederik S.

unread,
Nov 30, 2015, 3:57:11 AM11/30/15
to deal.II User Group
Dear Timo, dear Wolfgang,

Thank you for your responses. It's sadly not the answer I was hoping for, but I now know what I have to do to make the hanging node mesher work. 

Thanks again,
Greetings,
F.
Reply all
Reply to author
Forward
0 new messages