[dcyclo]

Dear Foamers,

Recently I was testing the overset mesh capabilities of OpenFoam-v2006 for a 2D vertical axis wind turbine case. My mesh consisted of: farfield mesh (level 0), refined mesh around the blades (level 1) and blade mesh (level 2).

I understood from technical description that in OpenFoam the background meshes are solved until the walls. Indeed, my “level 0” and “level 1” meshes were solved until the blades. Generally, I have found this a bit mean, but I also read that the solutions are corrected on the interfaces by interpolation (the well-known 4 cell-wise interpolation).

During the post-processing I have followed the suggested methods, starting from 33 mins of the following video:

https://www.youtube.com/watch?v=0mbmIOf3hRk

After using the tricks of slice, threshold, and translation of the meshes in the 3rd direction, I ended up a classical overset mesh solution with smooth interfaces between the mesh regions. Aerodynamic forces were also calculated correctly, using the obtained data from blade mesh (level 2).

Even though, the above written process is now straightforward for me and delivers the expected behavior from overset meshing, I have a huge concern regarding the residuals. In the residual monitoring, the residuals of all cells are listed. Obviously, the residuals of the level 0 but even level 1 meshes next to the blades are huge. As mentioned before, during the post-processing it became evident that these cells with the big residuals will be neglected from the final results. However, the simulation controls are based on residuals. I like letting the solver to use more nOuterCorrectors in the PIMPLE scheme if needed. During my sliding mesh cases, in general around 30 nOuterCorrectors iterations were needed for the PIMPLE scheme. Due to the residuals of level 0 and level 1 meshes however, in my overset calculations (no matter how many nOuterCorrectors I let the software to do) I never reach the prescribed convergence. It is also worth to mention that in example cases I saw a small number (1 or 2) for the nOuterCorrectors and I have not found any literature / video where people did discuss the convergence of overset mesh simulations.

After this thorough description I have the following question: does anybody know a trick to switch off the residual calculation for the background meshes? If you have any other solutions, I do also appreciate to hear them.



Thanks,
David