Time dependent PDE and mesh adaptation


439
views
1
11 months ago by
I want to solve a time dependent PDE in two space dimensions and I want to adapt mesh using my own criterion. Is there any demo example that explains how to do this ? All the examples I have seen are for stationary problems. 
Community: FEniCS Project
1
Be careful because FEniCS doesn't have coarsening. If you are modeling a wave phenomena, you might be refining regions that later on will be overrefined.
written 11 months ago by Miguel  

2 Answers


1
11 months ago by
If you have an instationary problem, you discretize in time and solve a series of stationary problems, so you can transfer the stationary case to this one.

You only have to make sure to adapt the problem and solver as well:
adapt(problem, refined_mesh)
problem = problem.child()
solver  = NonlinearVariationalSolver(problem.leaf_node())​
Its too bad we cannot do coarsening. Is there any development on this front in fenics ? Also, I am not using *VariationalProblem or *VariationalSolver. I directly construct matrix and solve it. Is it possible to adapt in this case ? Also in your tip above, how is the solution transferred to the refined mesh ?
written 10 months ago by Praveen C  
I don't know if there is any development....

Looking at the source code if you use adapt on a problem it adapts the corresponding Forms, Variables and Bcs.
You can do this manually in your case, i.e. first adapt your Forms and so on and afterwards build your matrices.

Again looking at the source code, inside adapt(problem, refined_mesh) the solution is also adapted, i.e. adapt(u, refined_mesh) is called internally.
And if I understand it correctly inside the adapt functionality a new (child) function on the refined mesh is created and the parent function is interpolated onto the refined mesh.

to be safe you can use

adapt(u, refined_mesh)
u = u.child()
written 10 months ago by Lukas O.  
1
11 months ago by
pf4d  
Each timestep, you can mark the mesh for refinement.  See here:

https://harishnarayanan.org/research/porous-flow/
Please login to add an answer/comment or follow this question.

Similar posts:
Search »