[Queiroz]

Hi guys,

hope you're all doing good. Just would like to ask you a simple question.

I've been working for several months in the generation of an "optimal mesh" aiming to solve my boundary-layer problem.

I used three types of meshes: (1) hybrid (tetra and prisms), (2) hexa without o-grid and finally (3) hexa with o-grid. I tried also several volume sizes and number of nodes.

To validate my simulations, I've been using some known experimental data.

The first mesh gave me acceptable results, the second one showed a flow pattern at the upper wall that was not true and the third one showed me, in comparison with the other two meshes, the best results.

When I showed a print screen of the three meshes used for a friend of mine, also in the Ph.D., he told me that the third mesh (the one that showed me the best global results) was totally wrong.

I've attached pics for each one of them. Could somebody tell me if he is right, what is the problem with this mesh? Thank you very much.

[bluebase]

Well, the third mesh is not very good because the change rate of cell size is too big at the interface between the o-grid and inner mesh (very small cells next to big ones). In fact you should try to make the third mesh looking more like the second one. With the current setting, there might be reduced accuracy for the flow in this interface, or it could lead to reduced convergence rate.

Another thing, the o-grid mesh cells are also a bit skewed. I suggest you to orient the cell edges to the inner mesh by setting different edge grading for the facing edges. Or you could use additional splits. The o-grid could look more like the boundary layer of the first mesh.

Also please be reminded that the "optimal" mesh size also depends on the meshing strategy itself. Check for grid independence for each of those three. Usually, unstructured meshes need more elements than structured ones to get grid independent solutions.