Does there exist a standard format for finite element meshes which is widely used in the industry?



"The nice thing about standards is that you have so many to choose from." -Andrew S. Tanenbaum

In my experience with using 3-4 different pieces of FEM software (and writing a few of my own) is that there is not a silver-bullet gold-standard out there. Some Creative Googling will give you some options, ranging in implementation from ascii-driven, XML based, to hdf5.

I would be really pleased to discover that I am wrong about this . . .

  • $\begingroup$ HDF5-mesh is dead, though that page cites some alternatives that use HDF5 as their back end. Xdmf uses both XML and HDF5. All the formats are bad, but some are less bad. $\endgroup$ – Jed Brown Mar 3 '12 at 5:01

Exodus II (utilizes NetCDF) is used by many ...


There is better pdf somewhere but I cant seem to find it right now

  • $\begingroup$ Thanks for the Exodus II link. I note that VTK 5.x has codecs for it, which is helpful in what we're doing. $\endgroup$ – Simeon Fitch Oct 10 '14 at 12:17

It all cooks down to

  • a coordinate table for all vertices in mesh
  • a connectivity table defining elements in terms of global numbers for vertices

which completely describes a given mesh.

It is easy to construct conversion scripts which can convert to other formats etc.

  • 3
    $\begingroup$ This is rather simplistic, there are also boundary sets (vertex sets and face sets defining Dirichlet and Neumann/Robin), subdomain/material markers, element type/topology, association with a geometric model, and often extra metadata to enable efficient parallel IO. If solutions are also stored in the same format, there is massively more relevant metadata. $\endgroup$ – Jed Brown Mar 3 '12 at 4:59
  • 1
    $\begingroup$ It used to be that most mesh generators output what Allan describes, but even for the simplest problem you need boundary markers. The most aggravating thing is that format such as ExodusII overcomplicate the problem. If you augmented the above with ONLY markers on arbitrary mesh pieces, that would be enough. $\endgroup$ – Matt Knepley Mar 3 '12 at 16:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.