[ziyasaydam]

Hi all,

I've been trying to model a boat in free surface flow, free to heave and sink in CFX 14. The solution is fine in steady state with k-e turbulence model.
When I use the results with the k-e model as initial conditions for a simulation with Reynolds Stresses Models, no matter what I do, I can't get it to work. After a few iterations, the resistance values tend to increase abnormally and the simulation ends with the linear solver overflow error. When I look at the velocity distribution with CFX-Post, I see extremely high velocity singularities at sharp corners.
So far, I have tried to reduce the time step, modify the mesh in corner areas, lowered the discretization orders, changed advanced turbulence settings, etc..
If anyone's got similar experience, I would appreciate some suggestions...

Thanks in advance...

Ziya

[ghorrocks]

My experience is that RSM is very hard to converge. It requires very high mesh quality and any other numerical instabilities (eg free surface modelling) can make it go unstable.

So I would recommend doing the best quality mesh possible. Other than that there is not much you can do.

Why are you using RSM on a boat flow? I would not imagine RSM would be a significant contributor to this type of flow.

[ziyasaydam]

It is mainly due to the streamline curvature effects existent in the flow and also
RSM was the only option amongst the ones I've tried that agree with experimental data...

Your point about mesh quality makes sense. I made RSM work in the naked hull condition but when the appendages were added to the model, the mesh quality was not as I desired and I get all sorts of problems.

[ghorrocks]

Sure there is streamline curvature in flow around a boat, but is it significant?

Also the SST model has a streamline curvature correction option. You may be able to use that and that is adequate - it certainly will be far easier to converge.

[ziyasaydam]

The problem with the SST model is the wall treatment. The mesh needs to be relatively coarse at the hull (100<Y+<300) to prevent air entrapment at very high speeds. In this case, I could not get a grid independent solution with the SST model with a coarse mesh. However, I will try the streamline curvate correction option...

[ghorrocks]

The wall treatment for SST and RSM is the same for that range of Y+. So I cannot see how an RSM model will help things, if anything it will make things far, far worse. You are less likely to get a grid independant solution with RSM compared to SST.

I suspect your problem is more related to the use of a coarse grid to stop air entrapment. My recommendation is you somehow need to stop air entrapment with a coarse grid (with an SST turbulence model).

[ziyasaydam]

Dear ghorrocks,

Sorry for staying silent for a while.
My concern with the SST model was the use of automatic wall treatment versus the wall function utilization in k-e and RSM models.
I have done some initial grid refinement studies and seen that SSt model is sensitive to first cell height in 100<Y+<300 range. In the meanwhile, results with k-e model with wall functions was insensitive to Y+ in that range.
Do you have any evidence that this may not be the case for the SST model?
In the meanwhile, I am running the simulations with SST model including the curvature correction option. Seems like over predicting drag about 10% when 100<Y+<300...

[ghorrocks]

Sorry for staying silent for a while.

No problem, there are plenty of other conversations around to keep me busy.

SST and k-e are using similar wall function approaches in the y+ range so the difference you are seeing is in the turbulence model, not the wall function. In general I would recommend SST over k-e, but there will be cases when the reverse is true.

10% accuracy on drag sounds pretty good to me. How accurate do you want it? To go more accurate than that I suspect you will need to do very careful sensitivity analyses on all important variables.