Learning Forum

[tim13]

Hi there,
I am doing an external aerodynamics simulation of a 3D model of a train under crosswind (cross-flow) effect. The mesh is hybrid and consists of a big hexahedral domain, and a smaller box inside the domain (around the train) containing tetrahedral and prism mesh that was generated using ICEM CFD. Minimum orthogonal quality is 0.05. Minimum quality criterion based on ICEM CFD is 0.1. Fluent case check has no recomendation in terms of mesh quality so far. Total elements are around 7M, and I am trying to keep it below 8M because of my hardware limitation.

Boundary conditions are as follows:
front: inlet 44m/s, 1% turb intensity, 0.05 length scale
left: inlet 20m/s, 1% turb intensity, 0.02 length scale
right: pressure out, 0Pa, 1% backflow, 0.02 length scale
back: pressure out, 0Pa, 1% backflow, 0.05 length scale
top: symmetry
bottom: moving wall 44m/s, specified shear=0

The calculation was first initialized by using first order k omega SST up to 500 iterations. After that, I switched to transient second order. I started to see message saying that turbulent viscosity limited to viscosity ratio of 1e+5 after around 250steps. The number of cells are increasing with every time steps, no sign of decreasing after around 200 steps. The time step size is 1e-3, and there are 30 iterations per time step.

Can anybody please enlighten me on how to solve the turbulent viscosity limited problem? Did I miss something in my fluent settings or my mesh quality might be the problem?

Thanks,
Tim.

[Kalyan Goparaju]

Hello Tim, 

The TVR hits the limits either when a mesh insufficient resolution is being used to calculated turbulence or in cases of large problems (in terms of length scales) such as flow over buildings. I am unsure if this case can be considered a large problem. Please see the following thread to get a sense of different strategies to use to resolve this issue. 

/forum/forums/topic/turbulent-viscosity-limited-to-viscosity-ratio-of-1-000000e-05-in-5-cells/

Thanks, 

Kalyan

[tim13]

Thanks for the reply, Kalyan.

So based on the link you've given, now I know that my mesh probably need to be refined. However, as I've mentioned earlier, I'm having a hardware limitation right now as I'm trying my best to put the cell count below 8 million (now it's 7 million already). How would you suggest to overcome the problem in case of the cell limitation?

I also don't understand the large problem term that is mentioned in the link you've given. How do we actually classify a problem as a large one so that we can alter the TVR limit? To add, my domain dimension is: length 120m, width 93m, height 30m. And the train inside the domain has a length of 20m, width 2.7m, and height 3.4m.

regards,
Tim.