[Puschkin]

Hello dear CFD Online Members!

Currently I'm facing a problem regarding the boundary layer mesh generation in ICEM. For my case I'm simulating a simle CD-Nozzle as a 2.5d Problem. Therefore I did the blocking, created a hexa-pre-mesh, converted it to unstruct mesh and finaly extruded it to 4°.



I use CFX to simulate my case and everything works like it should as long as I work with a smooth wall BC. I pre-calculated the wall distance and my y+ value (I'm using the SST-Model) is in the desired range of 0.4-0.7 over the whole domain with a "first layer thickness"/spacing of 1e-7 [m]. Everything converges fine and the results are very promising.


As soon as I add a wall roughness (3 micron) the y+ value goes through the roof. I get values of up to 50 and I'm not able to reduce the y+ value properly.



What I tried to overcome the problem:


Different "first layer thickness"/spacings (lower and higher than 1e-7 [m])
Different growth ratios between 1.05 and 1.2.


Lowering the "first layer thickness"/spacing helped to reduce the max. y+ value to 33 but min. y+ is still above 10. The problem now is, that as soon as I try to lower the "first layer thickness"/spacing any further, the mesh in the first layer(s) is getting distorted (see attached picture) and the extrusion of the unstructed mesh in the future steps fails.



index.jpg


So currently I'm limited regarding the "first layer thickness"/spacing (min spacing 2e-8 [m] before the problems occur). Since I'm trying to investigate the heatflux over the nozzle I would like to resolve the boundary layer correctly.



I read the manual of ICEM and ANSYS CFX and also the modeling guide and I'm aware of how ANSYS CFX models the wall roughness.



But sadly I'm currently out of ideas.
Maybe someone experienced a similar problem.
Any help would be appreciated!


Thank you!

[ghorrocks]

Quoting the CFX Solver Modelling Guide (section 4.2.4): "For rough walls, the logarithmic profile exists, but moves closer to the wall.", in other words, a given mesh will have a smaller y+ value with rough walls activated compared to smooth walls.

In addition, I think you need to consider whether integrating to the wall (ie, using a mesh with y+ ~1) is appropriate for a rough wall model. The turbulence models have wall functions, so why not use them - then a y+ ~1 mesh is not required. In addition, as the roughness you are modelling is likely larger than your first few element sizes you need to consider whether your mesh has got too fine.

Wobbly prism mesh - I agree, I also find that the mesh smoothing in ICEM can sometimes lead to wobbly mesh like what you show. Sometimes I get better results by turning prism mesh smoothing off and remeshing. Alternately, try a different prism mesh smoothing algorithm.

[Gert-Jan]

Alternatively, take a mesh that works fine and then in 'Edit mesh', split your inflation layer. As a result, the mesh won't get distorted.