interfacing different grid domains
 

Hi,

Has any one used the interface boundary condition in CFX?

need to know a few things,

1.Is this effective enough in transfering the grid values from one side to the other (eg structured grid side to unstructured grid side and vise versa) do you know how this is done?

2. Has any one done any tests in this regard to validate this option in CFX.

3. is there any lost of data,thereby effecting the solution,is yes then by what percentage.

4.does it give good solution for compressible flows also

Any suggestion is welcome.

Thanks for your time.

joseph

Re: interfacing different grid domains
 

I am using CFX5.5

joseph

Re: interfacing different grid domains
 

Hi Joseph,

The General Grid Interface (GGI) in CFX-5 is fully conservative, fluxes are transfered from portions of one face to portions of another (based on overlapping of subfaces) and all the terms of the equations remain active (i.e. coupled solution across interface too). It is NOT and interpolation at the boundary!

It does not matter what grid type you have on either side, and as long as your grid has roughly the same grid density, no information will be lost (a coarse grid cannot carry as much detail as a fine grid).

The implementation has been well tested and is good for all flow regimes, although you will have to wait until 5.5.1 to have multiphase work across it (anyone interested in running a pelton turbine?). The interface can cut right through a shock without any noticeable change to the shock wave (tested on a supersonic wedge). Again, as long as your grid density does not change (which does not imply that nodes have to line up), the solution will be unaffected. The only noticable difference is a slight increase in memory usage.

Regards, Robin

Re: interfacing different grid domains
 

1.Is this effective enough in transfering the grid values from one side to the other (eg structured grid side to unstructured grid side and vise versa) do you know how this is done?

Yes, it is very effective. Don't expect something for nothing though. The type of grid does not really matter. What is more critical is the change in grid density. If your grid density changes significantly across the interface it would necessarily effect your results in a bad way.

The appoach is similar to conserving flows into and out of a control volume, except in the case of the interface flow through the interface is strictly conserved. eg: mass flow in side 1 = mass flow out side 2, and the same applies for momentum, energy, etc...

The main thing to deal with is how to distribute the flows from one side to the other. This is done by finding the intersection of the two sides, which gives us the area fractions that each mesh face on the interface sees on the other side. It's a set of area fractions because the faces do not necessarily overlap.

This is all coupled implicitly as well, in that nothing is lagged at the interface. Control volumes on one side are implicitly connected, through the interface, to control volumes on the opposite side.

Raw GGI is pretty straightforward. You get caught up in the "modelling" aspects when you move to frame change models (eg: frozen rotor, stage, transient, etc..)

2. Has any one done any tests in this regard to validate this option in CFX.

I would ask CFX for this. GGI has been in use in CFX-TASCflow for many years, and the implementation is identical in CFX-5.5.

3. is there any lost of data,thereby effecting the solution,is yes then by what percentage.

No. Since the flows are conserved there is no loss of mass, or energy, or momentum, or whatever.

4.does it give good solution for compressible flows also

Yes, it works equally well on compressible and incompressible cases. MFR frame change models are a real workhorse for gas turbine design.

Neale.