[Ashkan Kashani]

Hi everyone,

I want to use CFX to simulate the transient 2D flow past a stationary rectangular cylinder crossing the free surface (see Figure attached). The main objective is to accurately predict the lift and drag. Assuming that the free surface dynamics would probably have a minor effect, I want to treat the free surface as a free-slip wall (rather than take an involved multiphase approach), thereby speeding up the simulation and avoiding convergence issues associated with the explicit modelling of the free surface. With this in mind,
(1) What is the best practice for the boundary conditions here? Particularly, how can I impose zero pressure on the free surface in order to get accurate force prediction?
(2) Is it possible to have a more efficient boundary conditions arrangement whereby the free surface dynamics could be resolved only over a small distance upstream and downstream of the cylinder?
Note the fact that the inlet and outlet must be far away from the cylinder leads to a long narrow domain; so the free surface dynamics is absolutely of no interest over the major part of its span.

I would appreciate any comments.

[ghorrocks]

1) To impose zero pressure at the interface just make it a pressure boundary set to zero pressure. You will probably have to use the entrainment option to handle the cross flow. But pressure boundaries do not work well with lots of cross flow like this so you will probably have problems with convergence.

2) This model should be tractable modelling it as a proper multiphase model with the free surface modelled properly. I see no reason to do dubious handling of the free surface, especially as the free surface is going to have waves propagate out a very long way (like the wake of a boat), so handling these with a boundary close by is going to be difficult.

Just model it properly in full free surface multiphase. It is not that expensive these days and you will be doing real results quickly rather than trying to get some contrived boundary condition to work.

A side comment: The gerris CFD code (open source) would do this model very quickly, nicely and more accurately than CFX. There is a quite a learning curve on gerris but it should do well on this model.

[Ashkan Kashani]

Thank you so much Glenn.

Originally Posted by ghorrocks

1) To impose zero pressure at the interface just make it a pressure boundary set to zero pressure. You will probably have to use the entrainment option to handle the cross flow. But pressure boundaries do not work well with lots of cross flow like this so you will probably have problems with convergence.

I have already imposed Opening with a relative pressure = 0 at the outlet (i.e. rightmost end of the domain). Is it mathematically/physically correct to set relative pressure = 0 at multiple boundaries? Could you please give further details?

Originally Posted by ghorrocks

I see no reason to do dubious handling of the free surface, especially as the free surface is going to have waves propagate out a very long way

Originally Posted by ghorrocks

Just model it properly in full free surface multiphase. It is not that expensive these days and you will be doing real results quickly rather than trying to get some contrived boundary condition to work.

I am not sure if I understand exactly what you mean here. Are you recommending multiphase modelling? Because I suppose you previously mentioned there's no point in dealing with the free surface dynamics in this problem.

[Gert-Jan]

If you want to close your eyes for the free surface effects, then apply a wall with free slip instead of a pressure opening at the top.

[ghorrocks]

..... which is then effectively a single phase 3D simulation with a symmetry plane. In other words, this is no different to a normal 3D, single phase, cylinder simulation; with the addition of a symmetry plane in the middle.

I would check your initial assumption that the free surface effects have no effect. For any moderate velocity or higher it will create a cavity in the water which will have a major effect. And if you decide it has no effect, then just look up the cylinder drag numbers from a fluids textbook as it has been measured to extreme accuracy many times by other researchers.

[Ashkan Kashani]

Hi all,

Could anyone please give me some tips on the appropriate turbulence models for this problem? Could it be simplified as laminar flow by any chance?

[Gert-Jan]

That depends on the Reynolds number......