[pi120]

Hi,

I am struggling to obtain a nice prism layer in proximity of the corner of my geometry. See the images attached to get an idea. Any suggestion on how to get better results?

On a side note, is it actually necessary to have a prism layer at all in the recirculating region? The whole area is a shear layer...

Thanks
Pietro

[LuckyTran]

You're trying to pack too many layers into too small a distance relative to the surface mesh. It's going to retract the prism layer thickness to maintain mesh quality, it can't because that thickness is less than the minimum thickness (another setting). Therefore it retracts it to 0.

Reduce the number of layers or allow the mesh to be smaller (i.e. increase cell count).

More important than y+, you're always going to want prism layers next to walls. Try even an inviscid case without prism layers and you'll see right away how important they are and it has very little to do with y+.

[pi120]

Thank you.
Could you explain why you need a prism layer for an inviscid simulation? I always thought of the prism layer as the area where you try to capture the physics of the boundary layer. As an inviscid flow has no boundary layer, I don't see why you'd need it.

Just as a general thought, I noticed that using a prism layer in the recirculating area leads to areas where the velocity is perpendicular to the prism layer cells (see attached image), which should affect negatively accuracy and residuals due to their low aspect ratio. But I guess that's something you have to accept when you want to model wakes.

[LuckyTran]

More important than capturing boundary layers is capturing basic kinematic conditions. The velocity vector next to walls needs to be parallel to the wall and not into the wall or you will have all sorts of problems interpolating/extrapolating the solution onto the wall boundary. Your pic clearly demonstrates this happening and why it is so desperately needed. Try repeating it without prism layers and see what happens to flow near solid boundaries.

[pi120]

Ok, so you need the prism layer to enforce the no-penetration condition. Makes sense, thank you