Learning Forum

[Antang]

My colleagues have always insisted on making structured meshes which I understand can be computationally easier to make and more standard.  However, I'm given a unique multi-element airfoil geometry that I simply cannot make with a structured mesh (with ANSYS Mesher) but I can draw an offset "bubble" around it and add an appropriate inflation to get an appropriate y+ value with 20 layers with 1.1 growth rate.  In terms of accuracy, If I use a K-omega SST turbulence model (or any other good model), will I still get equally good results compared to if I got a structured mesh, assuming I reached the same convergence?  Or will be difference be the convergence such that with a structured mesh I would get a better convergence to say it is 99% accurate/converged with that model while an unstructured mesh + inflation might be 95% accurate/converged with that same model?

[Rob]

Pretty much everyone uses unstructured meshes now, and the difference is negligible. Try both on a simple model and compare the results. 

[janetj]

Does this apply on 2D airfoil simulation as well? Say I am not using face meshing or my face meshing result is not too great, will the CL and CD results have a negligible difference comparing with the experimental results?

[DrAmine]

Applies 2d and 3d. For boundary layer it is always recommended to use structured say parallel mesh to the contours to easily capture gradients normal to the walls and the flow has a predominant direction. You need here not to confine the development of boundary layer by bad mesh. In the bulk and in case there us no real dominant flow direction which a hex or quad cell aligns to there is no superiority only in terms of cell count.

More details can be found for example in ERCOFTAC best practices.