RBF Mapping diverging

I believe I’ve read every post on here related to RBF mapping. I also read:

[1] Chourdakis, G., Davis, K., Rodenberg, B., Schulte, M., Simonis, F., Uekermann, B., Abrams, G., Bungartz, H.-J., Cheung Yau, L., Desai, I., Eder, K., Hertrich, R., Lindner, F., Rusch, A., Sashko, D., Schneider, D., Totounferoush, A., Volland, D., Vollmer, P., and Koseomur, O. Z., 2022, “PreCICE v2: A Sustainable and User-Friendly Coupling Library,” Open Res. Eur., 2, p. 51.

I’ve arrived at RBF mapping after attempting nearest-neighbor and nearest-projection for this case and have found them to prescribe deformations that my fluid solver cannot handle. My last chance is RBF, and I just want to make sure I’m understanding everything correctly and have exhausted every option before I give in and remesh the solid domain. I’m attempting to use a fluid mesh and solid mesh provided from a workshop to conduct FSI analysis of a wing mounted in a wind tunnel. Here is an image showing the .vtp exports of the fluid mesh (gray) and the solid mesh (purple):

The light blue sphere represents my defined support radius (1.5), and if we zoom in we can see it captures about 5 solid mesh nodes in this location:

My issue arises on the read of the displacement data from the solid mesh (purple) to the fluid mesh (gray). I’ve varied the support radius from as low as 0.5 to as high as 20 and always found the RBF solve to diverge. I’ve also tried rbf-thin-plate-splines, rbf-gaussian, and rbf-compact-polynomial-c0 without success.

Viewing only the top surface of each mesh highlights that the provided solid and fluid domains do not align very well.:

Is it possible that a support radius of 1.5 is too small to create an interpolant because of the difference in the Z location of these points, but then if I increase the support radius size the RBF matrix becomes ill-conditioned as it pulls from a larger number of solid mesh points?

Here are some relevant files:
precice-config.xml (3.6 KB)
raven.out.log (2.4 MB)

For reference, RavenCFD is our in-house CFD solver. Velodyne is our in-house FEA solver.

Any insights will be most appreciated.

Somehow this went under our RADAR. Sorry about that!

@DavidSCN @uekerman do you have an idea here?

According to the error message, there is convergence in the linear solver, but the solver fails to reach the prescribed tolerance within the max iterations. You could just lower the tolerance (maybe 5e-8 or so) to make it work. Not sure about the resulting accuracy. In this regard, I would rather recommend to upgrade preCICE to v3.x, as this seems to be preCICE version 2.x.

This topic was automatically closed 3 days after the last reply. New replies are no longer allowed.