[nmartin9319]

Hi!

I have to calculate the flow field in a stirred vessel (15 mm radius, 15 mm height) where the blades (10 mm radius) at the top of the geometry rotating with 3000 RPM.

It's a transient simulation, and until now I've tried to solve it with SST, since the Reynolds number is around 150000. The convergence is good, but if I want to keep the abs RMS at 1e-5 in every variable, then I have to run it with small stepsize (~1e-6 s) for about 5 seconds. Unfortunately I don't have that much time.

First it seemed like an idiotic idea, but I've tried to solve it without any turbulence modelling, and it worked. With 1e-3 s stepsize, the abs RMS in the velocities decresed to 1e-6 and to 1e-4 in the continuity in 10-20 iterations. The results also seem valid.

My question is that why the laminar model working in an obviously turbulent case. My mesh consists ~500k elements, so I doubt that I'm in the DNS regime. Should I believe my results? (I'd love to, but not sure)

Thanks in advance

[FMDenaro]

Originally Posted by nmartin9319

Hi!

I have to calculate the flow field in a stirred vessel (15 mm radius, 15 mm height) where the blades (10 mm radius) at the top of the geometry rotating with 3000 RPM.

It's a transient simulation, and until now I've tried to solve it with SST, since the Reynolds number is around 150000. The convergence is good, but if I want to keep the abs RMS at 1e-5 in every variable, then I have to run it with small stepsize (~1e-6 s) for about 5 seconds. Unfortunately I don't have that much time.

First it seemed like an idiotic idea, but I've tried to solve it without any turbulence modelling, and it worked. With 1e-3 s stepsize, the abs RMS in the velocities decresed to 1e-6 and to 1e-4 in the continuity in 10-20 iterations. The results also seem valid.

My question is that why the laminar model working in an obviously turbulent case. My mesh consists ~500k elements, so I doubt that I'm in the DNS regime. Should I believe my results? (I'd love to, but not sure)

Thanks in advance

You are solving a coarse DNS, also denoted as LES no-model. But that can have some physical meaning only if you run a 3D case and the grid is quite refined close to the wall. Depending on the action of the local truncation error, if it dissipates by means of the numerical viscosity, this approach is also denoted as Implicit LES (ILES).

[nmartin9319]

Originally Posted by FMDenaro

You are solving a coarse DNS, also denoted as LES no-model. But that can have some physical meaning only if you run a 3D case and the grid is quite refined close to the wall. Depending on the action of the local truncation error, if it dissipates by means of the numerical viscosity, this approach is also denoted as Implicit LES (ILES).

Thank you for your quick answer! Yes, it's a 3D model. I've attachetd two images of my boundary mesh. So there might be a chance, that my results are not completely wrong?
Mesh1

Mesh2

[FMDenaro]

Originally Posted by nmartin9319

Thank you for your quick answer! Yes, it's a 3D model. I've attachetd two images of my boundary mesh. So there might be a chance, that my results are not completely wrong?
Mesh1

Mesh2

The mesh appears not enough smooth in the refinement..however you should also compute the y+ you get close to the wall. But I suspect that with only 500K grid points you cannot have any physical results in terms of viscous action over the vessel. What about your scheme?

[LuckyTran]

Of course laminar model will work, the computer just solves whatever equation you tell it to. Not using a turbulence model basically just means you no longer have an eddy viscosity.


You'd have to check your grid resolution to see if you are like a DNS or not. But given that you are able to take much bigger timesteps with the laminar model, I'd say probably not. Otherwise, you'd be having the opposite result.