For structural modeling of very large geometries with allot of components, how stable is Fenics?

Question

I need to do structural FEM for large geometries which will include a high amount of components each to be meshed separately and boundary conditions defined in different ways for these meshes. I am concerned as it seems to me Fenics is more focused on fluid modeling as well as the shear scale of modeling that will have to be completed. Also it seems to me that I will need to write my own MPI parallelization if I would like to use clusters?

Answer 1

FEniCS is very stable at large scale with MPI parallelism. I have run simulations with up to 24000 processes, and 12.5 billion degrees of freedom.

What is less well tested is the behaviour with disconnected meshes, but it is not, in principle a problem. If you have separate components, with separate meshes, why do they need to be modelled together - what is the nature of the coupling between them?

Comments

Hi Chris, I want to investigate the influence of these components on one another (thus each mesh will influence the boundary conditions of the next), I am considering various modeling techniques for this, among other using cohesive zone elements to model the interfaces or alternatively to build into the code itself a mathematical model simulating the "porosity" created by using bricks instead of a solid wall for example. It seems another problem I might encounter is the lack of cohesive elements in Fenics.

---

Hi,
What you are describing is usually called a "contact problem", and is an area of active research in FEniCS, especially in parallel. There are no examples at present, but we are likely to add some in the next year or so.

---

Thank you for the information I am still in the process of figuring out how in detail I will need to model the contact areas.