Learning Forum

[jingyuliu]

Hi,

I am running the CFD in ansys fluent. The model geometry is as follow.

I’m working on a model in ANSYS Fluent where an interface (highlighted by a red dashed circle) consistently causes instability during simulations. I created the mesh with 13 boundary layers, setting a minimum element size of 0.01 inches and a maximum of 0.9 inches. I'm using a steady-state k−ϵ turbulence model for the simulations. With a convergence criterion of 1e-5 for continuity and 1e-6 for x, y, and z velocity components, the model successfully converges when water (liquid) is used as the working fluid. However, the simulation fails to reach this convergence criterion when using oil (liquid), which suggests a potential mesh quality issue. Currently, my mesh requirements focus on maintaining skewness below 0.8 and orthogonal quality above 0.2. Despite these efforts, the mesh quality does not seem adequate for stable simulation results with oil.
If i use less boundary layers, like 3 or 5 boundary layers, it is fine. Questions:

1. Are there additional mesh quality metrics or thresholds (e.g., aspect ratio, cell quality) that might be relevant for improving stability, especially for oil simulations?
2. Are there specific guidelines or best practices in mesh generation for multiphase simulations or for fluids with different properties (e.g., viscosity, density)?
3. Would refining mesh near the interface or adjusting the boundary layer thickness be beneficial in this context?
4. Is ICEM useful for this geometry problem? 
5. Is there any way to improve the mesh after the mesh generation?

Any insights or recommendations on enhancing the mesh for improved stability in this scenario would be greatly appreciated.

best regards,
Jingyu Liu

[Rob]

1. No, but remember inflation meshing is also a high aspect ratio if the solution changes along the cell. 
2. Yes, use inflation with extreme caution - see above. 
3. Possibly. That's one of the reasons we have dynamic adaption with pre-built criterion for VOF. 
4. No, but also avoid poly-hexcore in Fluent Meshing as the jump in cell size in the core mesh can be problematic if it coincides with the free surface. 
5. Yes, but I recommend doing it properly in the first place. 

With multiphase instabilities may be real and a result of the flowfield.