[Daaman]

I am performing a steady-state RANS simulation of a 6-blade propeller using a frozen rotor approach and periodic BC, by modeling a single blade. I have used ANSYS Meshing to perform meshing of my domain using unstructured mesh, with prism layers along the propeller blade and the propeller spinner. I have created 5 meshes based on the mesh refinements, by varying the volume and surface mesh settings while keeping the settings of the prism layer constant.

As shown in the figure, each data point represents the figure of merit for each mesh, with coarser to finer from left to right. It turns out that the coarser mesh i.e. Mesh 1 tends to have better agreement with the experimental value as compared to the finest mesh i.e. Mesh 5. (See Figure) I have tried changing the factor that is multiplied to set the mesh setting, but the trend remains the same. What could the reason be? Thanks in advance!

[jojoFD]

Hi,

What I remember well from my CFD classes is that you shouldn't try to make the mesh fit your experimental results. If something is wrong when you refine, then there is definitely a problem somewhere (sorry for stating the obvious Can you give more information on your settings please?

[LoGaL]

Your solution is still not mesh independent as I can see, because you should reach a zone where your curve is flat

Reach first that zone, then if you don’t match yet, start doubting the modelling or boundary conditions. The ratio with which you increase the element size is close to irrelevant

Also, are you kidding me? Is it really a 1% relative error? Are you chasing butterflies?