[sasanghomi]

Dear friends,

In this thread, I want to ask you about the quality of boundary layer mesh precisely.
In fact, I have never seen people who talk about the thickness of boundary layer mesh that are generated mostly by the inflation feature in Ansys Meshing on surfaces. So, please let me know your opinions about the following point;

According to Ansys user guide, the minimum number of mesh layers inside the boundary layer should be 10.

However, I have seen many cases that numerical people, metaphorically speaking, just justify the number of inflation around 10 and they care about the first layer thickness in order to preserve Yplus value.

Being more elaborate, imagine that there is a simulation in which the value of Yplus is acceptable and the number of cells in boundary layer generated by the inflation feature is 10. I believe that there is another point which needs evaluation. Are we sure that the boundary layer in covered completely with the boundary layer mesh? Absolutely Not.

My strategy is as below;

At first, we must simulate the fluid flow with a primary boundary layer mesh. Afterwards, as per the value of yplus and the first layer thickness, we can calculate the thickness of boundary layer in different parts of the model. So, in the next simulation we must assure that boundary layer has been covered completely. (When the Yplus value is around 1000, we have reached the edge of boundary layer almost.)

Also, I think the above-mentioned point is more than the concept of grid dependency and it needs a specific attention.

I would appreciate it if you clarified your viewpoints.

Best Regards
Sasan Ghomi

[karachun]

Here is an article about visualization of boundary layer in Fluent and CFD-Post. It is in Russian language but you can translate it via google.
http://www.cfd-blog.ru/visual-ansys-cfd-post/
Maybe it is possible to make such visualization in CFD-Post and CFX results too, and note perform runs on coarse mesh.
If you calculate some problem where boundary layer is crucial, like airfoil, then using coarse mesh with y+~1000 you can get results that significantly differ from results on converged mesh. Therefore, all that efforts on coarse mesh will be useless.

[ghorrocks]

Whether your y+=1000 results are useful depends on the situation. There are many flows where the boundary layer does not contribute much - it would be fine in those situations. Both for flows where accurate boundary layer resolution is critical - it is likely to be useless.

So you really need to do a sensitivity analysis of your simulation to mesh resolution. That gives firm, reliable numbers to guide development of a simulation strategy.

Also have a read of journals and textbooks on CFD accuracy. This FAQ links to an excellent article on the issue: https://www.cfd-online.com/Wiki/Ansy...publishable.3F, and the work of Celik and Roache on CFD accuracy is highly recommended reading.