[songkai]

Hallo Guys,

i am now working about my master thesis about the Simulation of the fluid in cylinderroom with Star CCM+. an easy question: if i choice DNS and laminar model in pysical continua, then need i to be meshing with prism layer or not?

thanks

[LuckyTran]

What is your core mesh type? Tets?

Prism layers are always a good idea, especially on tetrahedral-based meshes because it helps calculating the wall-normal and wall-parallel gradients. When you don't have prism layers, you often end up with velocity vectors that point into the wall. If you have a mesh that wraps nicely around the body then you don't explicitly need prism layers because you have these nice shaped cells already.

[songkai]

Originally Posted by LuckyTran

What is your core mesh type? Tets?

Prism layers are always a good idea, especially on tetrahedral-based meshes because it helps calculating the wall-normal and wall-parallel gradients. When you don't have prism layers, you often end up with velocity vectors that point into the wall. If you have a mesh that wraps nicely around the body then you don't explicitly need prism layers because you have these nice shaped cells already.

thanks and i would like tets. the model is very easy just to simulation the fluid in cylinder and i need to focus on the layer (surface) on the combustion chamber wall. i have finished y+=60 with RANS and also with prism layer but i am not sure, should i need to set prism layer under y+<1, if i set the mesh very small.

thanks

[me3840]

If you are doing DNS you may not need prisms depending on what the flow looks like. How much do you expect the kolmogorov scale to change in the domain?

I would avoid using tets for DNS, they're likely too diffusive. I would make a structured grid for that.

Though I'm not really sure what discretization schemes are available that would be useful for DNS in STAR-CCM+...it would be best to use something not diffusive and high order, which is not really available. Unless in newer versions a DNS model exists?

[songkai]

Originally Posted by me3840

If you are doing DNS you may not need prisms depending on what the flow looks like. How much do you expect the kolmogorov scale to change in the domain?

I would avoid using tets for DNS, they're likely too diffusive. I would make a structured grid for that.

Though I'm not really sure what discretization schemes are available that would be useful for DNS in STAR-CCM+...it would be best to use something not diffusive and high order, which is not really available. Unless in newer versions a DNS model exists?

Thanks
the flow is a cycle flow in the cylinder depending on the strokes. the kolmogorov scale need to be calculted depending on the inlet velocity?