Fluids

Fluids

Running a two-way intrinsic FSI simulation with non-conformal meshing.

    • johnhavenar
      Subscriber

      I am attempting to model the aeroelastic response of an aircraft wing using the Structural model within Fluent. I have done the same task for many configurations before using the Fluent + Mechanical + System Coupling workflow, but would like to switch to working strictly in Fluent.

      The polyhedral meshing for the fluid domain reduces my solution time by 50% and matches experimental data much better than the tet or hexcore methods, so I am trying to use it in the fluid domain and then create a non-conformal interface between the fluid and solid domain. According to page 2812 of the 2021R1 Fluent User's Guide, "Solid-solid non-conformal interfaces are not supported; only fluid-solid interfaces are supported," so I should be fine.

      The simulation and dynamic meshing works fine with conformal pure tet meshes using wall-wall-shadow pairs, but suffers accuracy for Mach numbers 0.90-0.98, so I am still trying to make use of the polyhedral scheme. When using the non-conformal interface however, the error messages in the attached photo occur and the software crashes. The first time step solves as expected, but then once the dynamic meshing begins (diffusion smoothing) it fails. It's a bad crash, too. No meaningful messages, and the parallel application simply dies until the interconnect times out. Additionally, when reopening the case file saved at the start of the two-way analysis, if I display one of the walls created as part of the non-conformal mesh interface, the application dies the same as before.

      Any help would be greatly appreciated. Thanks.

    • Stephen Orlando
      Ansys Employee
      Hi John This looks like a bug that was observed previously with non-conformal meshes. Are the fluid and solid mesh elements of similar size at the FSI interface? Try changing the mesh size at the FSI interface.
      Steve
    • johnhavenar
      Subscriber
      Thanks for getting back to me.
      The fluid surface and solid surface were both meshed using a minimum size of 0.5 mm, a maximum of 5.0 mm, and a growth rate of 1.2. The difference was that the fluid volume used poly-hexcore and the solid volume was pure tetrahedral.
      You mentioned that this was observed before, but I haven't been able to find anything on it. Could you possibly link me to the source?
      I also tried creating a conformal pure tetrahedral mesh and exporting the fluid and solid to separate mesh files, loading the fluid domain, converting it to poly while preserving boundary layers (so it should still be conformal), and then appending the solid mesh to the fluid mesh and fusing them together. The fusing was successful and a wall - wall-shadow pair was created, but the analysis failed once iteration began.
      John
    • Stephen Orlando
      Ansys Employee
      detailed information about bugs isn't publicly disseminated. I think you'll need to stick with the conformal mesh in this situation.
      Steve
    • ujjsf
      Subscriber

      I'm also attempting to run a FSI simulation in Fluent with a non-conformal mesh interface and encountering the same 'cannot locate node XX' error after a few time steps. I got only hex cells in my grids, the solid mesh is a lot coarser though and the nodes of the fluid mesh and the solid mesh are not matching - which I would like to retain like that. With a conformal interface (meshs have the same coarseness, nodes match) there is no problem at all but the solid mesh is way too fine now...
      Did you maybe find a solution for the problem besides using a conformal mesh?

      Thanks in advance!
      ujjsf

    • johnhavenar
      Subscriber
      Hey Unfortunately, I was not able to find a solution which used a non-conformal interface. I eventually bit the bullet and used a conformal interface and tweaked the growth rate of the solid mesh to save on cell counts. I'm not sure what you're modeling, but I found that the solid domain cell count did not strongly influence the compute time required, since the fluid cell count dwarfs the solid cell count for my case.
      Thanks John
Viewing 5 reply threads
  • You must be logged in to reply to this topic.