URIS_displacement appears incorrect

[aabrown100-git]

Description

When running an FSI simulation with URIS valves, the output array URIS_displacement in result_uris_<valve_name>_xxx.vtu seems incorrect. When I visualize the in Paraview by warping by URIS_displacement, the valve motion does not follow the fluid mesh motion.

I encountered this when running a pretty complex left heart FSI simulation with valves. Here is the valve position when warping by URIS_displacement (the timestep is in the middle of the simulation when the fluid mesh has deformed significantly)

And here it is when warping by the fluid mesh displacement sampled onto the valve

In both pictures, the elements around the leaflets indicate where the valve resistance is being applied, as measured by the signed distance field (SDF). Apparently, the SDF field is correct, which is what really matters. So this seems to only be an outputting bug; the actual simulation results (i.e. fluid velocity, mesh displacement, etc.) are probably fine.

Reproduction

Run an FSI simulation with RIS valves. The simulation should be designed so that the fluid mesh moves significantly over the course of the simulation.

I want to provide a minimal example, but don't have the bandwidth right now.

Expected behavior

The array URIS_displacement should be computed correctly so that when warping the valve surfaces by URIS_displacement, they follow the fluid mesh as it deforms over the course of the simulation.

Alternatively, we can just remove URIS_displacement as an output array. It's very easy to compute the valve surface motion as a post processing step, using Paraview or PyVista. As far as I can tell, the valve surface displacement is only computed to be output as URIS_displacement. It isn't used anywhere else in the URIS code and doesn't affect the solution. Removing this would also allow us to eliminate several rather complicated function (uris_update_disp, uris_find_tetra).

Additional context

No response

Code of Conduct

• I agree to follow this project's Code of Conduct and Contributing Guidelines

[aabrown100-git]

@hanzhao2020 Do you have any thoughts about this?

[hanzhao2020]

@aabrown100-git The current implementation calculates the URIS object displacement Yd looks reasonable. An example would be helpful to reproduce this issue. If the example is too much work, we can try the post-processing only step.

[aabrown100-git]

@hanzhao2020 Sorry for the delay. Here's an example (pipe_uris_3d.zip) modified slightly from the pipe_uris_3d test case.

I add a script called compare_uris_displacement.py that compares the original URIS_displacement array output by svMultiPhysics and a URIS_displacement_recomputed calculated by the script by sampling the fluid mesh displacement onto the valve mesh. If you run it, it outputs some measures of the difference between these two arrays. You can see they are not the same. You can also open valve_mesh_with_comparison.vtu in Paraview and warp by the two arrays to visualize the difference (you'll have to use a scale factor of around 10 to see the difference).

For this example, the difference between the two arrays is small, but I think it is real. Also, for my FSI simulations, the difference is much more significant.

Again, one simple solution is to just remove the calculation of URIS_displacement from svMultiPhysics. It is very simple to just sample the Displacement array from the fluid mesh as a postprocessing step, exactly as is done in compare_uris_displacement.py.