Hej,
I am interested on the gas motion due to a cyclic movement of a wall.
In order to keep things simple I run a simple case in 2D.
So it is defined:
- Navier-Stokes
- Heat Tranfer
- Mesh Update
With small wall movements it smoothly runs; however when the mesh is deformed beyond a certain point it crashes. In other software it is possible to input different meshes in order to jump form one to the other to avoid over-deforming it. May be How generate a new mesh when it becomes too contrained?
I have been reading everything about "adaptive mesh" in Elmer however I do not succeed.
How can I define a mesh (or a joint of meshes) which can handle such a deformation?
I attach my working files.
Thanks in advance,
Mesh Update - Higly deformed mesh; maybe an adaptive mesh?
Mesh Update - Higly deformed mesh; maybe an adaptive mesh?
- Attachments
-
- MovWall2D.tar
- MovingWall2D
- (720 KiB) Downloaded 277 times
-
raback
- Site Admin
- Posts: 4862
- Joined: 22 Aug 2009, 11:57
- Antispam: Yes
- Location: Espoo, Finland
- Contact:
Re: Mesh Update - Higly deformed mesh; maybe an adaptive mesh?
Hi
Unfortunately there is no remeshing option out of the box. There is the adaptive machinery that has many of the needed building box. However, it does not have the ability to deform the geometry that is sent to the mesh generation software. The adaptivity of Elmer has not entirely kept up with pace with the other developments. Generally the adaptivity in Elmer is limited to a few equations and parallellism with MPI is not supported. Also there are not that good 3D mesh generators available in 3D to make the desired meshes, in 2D the problem works better.
Your case would require some tweaking of the code. There chunks of code from Adaptive.src could be used but also a strategy to make the mesh deformation compatible with multiple meshes would require some attention. If there would be N meshes there should also be an offset vector with size N-1 that would tell how the displacements in various meshes are considered, and and decide when to switch between different meshes. One could start with simple translations.
If the meshes just deform around the corners a possible solution is to make the mesh poisson ratio larger around these problematic points. One can write a UDF where the rigidity depends a from minimum distance from the selected points by some decaying function.
-Peter
Unfortunately there is no remeshing option out of the box. There is the adaptive machinery that has many of the needed building box. However, it does not have the ability to deform the geometry that is sent to the mesh generation software. The adaptivity of Elmer has not entirely kept up with pace with the other developments. Generally the adaptivity in Elmer is limited to a few equations and parallellism with MPI is not supported. Also there are not that good 3D mesh generators available in 3D to make the desired meshes, in 2D the problem works better.
Your case would require some tweaking of the code. There chunks of code from Adaptive.src could be used but also a strategy to make the mesh deformation compatible with multiple meshes would require some attention. If there would be N meshes there should also be an offset vector with size N-1 that would tell how the displacements in various meshes are considered, and and decide when to switch between different meshes. One could start with simple translations.
If the meshes just deform around the corners a possible solution is to make the mesh poisson ratio larger around these problematic points. One can write a UDF where the rigidity depends a from minimum distance from the selected points by some decaying function.
-Peter