[jbrydg01]

Hello!

I'm running a 3D simulation of air flow around a flat plate with an aspect ratio of 1.6. I am doing this using simplefoam, and because my boundary is rectangular and my plate is perfectly perpendicular to the oncoming flow, I am using blockMesh with the intention of getting the highest quality mesh and control. I've attached the blockMeshDict file if more information is needed, but to summarize:

• The boundaries are 25h in the x-direction, and 16h in the y and z-directions.
• The plate is located 5h back from the inlet.
• The plate is specified as no-slip and the side, top and bottom walls as symmetry. The inlet has a constant velocity of 1.01m/s. For pressure, the inlet is zeroGradient and the outlet is a fixed value of 0. The plate is zeroGradient.

Because of limitations to the cell size in the x-direction, I switched from modeling a 3D plate (with a specified thickness) to a 2D baffle using the topoSet utility and the createBaffles utility. I've included both the files.


What I'm finding is that when I have a mesh at around 1-2 million cells, my simulations will run and converge. While my coefficients of drag and recirculation length do not entirely match with those of the literature I'm comparing to, that is a different issue. However, when I increase the grid resolution higher than this, my coefficient of drag will appear to converge for the first 50-100 iterations, then will begin to oscillate and explode (just gets to be a ridiculous number).


As another note, I've found that what tends to help is if my transitions from one block to another have cells be as close to the same size as possible (hence the slight grading in the blockMeshDict). I've also found that having matching cell sizes in the x, y and z directions have helped, but is not essential. Any comments/validation on these observations would be greatly appreciated!


I was hoping to get some feedback on why this may be happening. Obviously I am solving some large gradients around the plate and to my knowledge RANS simulations do not handle sharp corners very well. I've tried looking into evidence of openFOAM not handling 3D simulations at high cell counts very well, but I can't find anything that explicitly supports this.



Thanks for your feedback!


Jesse