"greatly simplified geometry"

Nah, I don't overestimate in general. I only underestimate myself in general (but nothing to do with the effect that you are the guy 😎)

Ah... Mmm yeah true but I thought understand that actually it was the god geometry but not repeated

Hello, does anyone have a viscoelastic RANS turbulence model?

No

hey otaolafr, what software is that? is it SALOME?

if so is there any tutorial around available for block-based meshing in SALOME that you would recommend?

waaaait a minute I just googled for literally 20 seconds and there is a block-based mesher module in SALOME?????

hexablock?

I have been using SALOME for literally 3 yrs and never knew

Because I'm planning to use the kOmegaSSTDES, SpalartAlamarasDES and WHALE turbulence model for a unsteady simulation, I'm expecting using the results of a RANS turbulence model as a IC. I'm wondering, is it preferable to use the corresponding RANS model for the IC, or it doesnt matter at all? For instance, in the case of kOmegaSSTDES unsteady turbulence model, Should I use the result of the RANS-kOmegaSST turbulence model as a IC? Same thing with the SplartAlamarasDES, should I use the RANS-SpalartAlamaras results as a IC? If don't, I'm expecting to use the same RANS results as a IC for all the unsteady simulation. What is your suggestions/thoughts about this topic?

You have to use the same family (Spalart-Allmaras will require initial conditions for nuTilda which you won't get from other models), the same applies for kOmega.

You could use kEpsilon and kOmega interchangeably and calculate epsilon or omega from the other set. But here you'll have a low-Re mesh and kEpsilon doesn't like that

Thank you very much for replying! This is exactly I was thinking of....Some fields are specific to the turbulence model itself. But I was still wondering if it has a trick to save simulation time by using the same RANS results for different unsteady turbulence model. On the other hand, I'm still wondering which one should I use as a IC with LES turbulence model like WHALE or Smagorinsky for instance. If I remember well, these type of turbulence model use U and p only. That means, any RANS model could be use for U and p right?

Yes, you're right, for LES models that only work with p and U, you can use any RANS model.

@t_bo99, are you planning to run steady-state simulations and get their results as your initial conditions? I don't know if this makes sense but you could run for example, k-omega-sst to get the results of p and U. Then you may use these p and U results to get say k and omega by running another but "sort of converged" simulation. Please correct me if this idea seems implausible.

If I understand what you saying is....instead of starting a new RANS simulation from scratch, I could use U and p results from a previous RANS simulation to resolve another sort of RANS simulation?

Yeah that's what I had in my mind. Only things to solve for would be your turbulent unknown fields. I don't know if this approach makes sense and correct, I never ran LES before but I am planning to run one.

I would use potentialFoam instead as a IC for each different RANS simulation. Once I got the results from these differents RANS simulation, like @kandelabr said and answered well my first post, the goal is to use these results as a IC with the same family/correspondant unsteady turbulence model....

hey LES experts, how sensitive is LES to high growth ratios?

Oh gosh, How I love these questions: very sensitive, or not at all 🙂

suppose it's very sensitive if we're talking about growth ratios in the direction of flow?

That depends on soo many things and depend on whether you are talking about wall-modelled vs Resolved, close to the wall, explicit/implicit filtering, ILES/LES, FVM/FEM/FR/SEM, type of flow... Your question is too broad for me, or anyone for that mattter, to give you a serious answer.

You need to be specific, VERY SPECIFIC, on what you are asking

understood! So I'm simulating gas flow in a cyclone separator

I'm doing wall-modeled, explicit LES with FVM

the flow is very anisotropic - small gradients in tangential and axial direction, larger gradients in radial direction (both pressure and velocity fields)

here's the mesh:

and as you can see, there are jumps in cell size in the core, and then again under the vortex finder wall (beneath the open space)