problem with GGI interfaces
 

I similated a flow threw a serie of rectangular pipes including different meshes. I set GGI interfaces between those meshes and the simulation converged. The thing is that when I checked the results, I observed great difference of energy level on each side of the interfaces, even if it's physically the same plane i'm checking. It seems that there is a «numerical» head loss across the interfaces. Am I wrong? Why does it occurs? Can I solve that?

Any help would be very apreciated. Thanks

strider

Re: problem with GGI interfaces
 

Hi Strider,

The GGI interface will conserve mass, momentum and energy. However, other quantities, such as the total pressure, may change due to the change in your mesh resolution. You would incurr the same losses if your mesh was fully conected but still reduced the mesh density.

What exactly are you checking and what energy equation are you solving (ie Thermal or Total Energy)?

Regards, Robin

Re: problem with GGI interfaces
 

thank you Robin,

it is indeed the mass flow average total pressure that drops across the interfaces. I don't get how energy is conserved but not this variable? The goal of this first simulation is just to simulate the flow in the rotor-stator generator so that the mass flow at different points collapses the experimental data. So it's isothermal for the moment. The rectangular pipes I told you about are the «holes» in the rotor and in the stator that permit the cooling. i think i'll have to check my meshes on both sides of the interfaces, but at first sight the resolutions seem similar. At the stage interface, the pressure is almost the same on each side of the GGI, but the total pressure goes from -4000 to 5000 Pa.

by the way, do you think that imposing a porous volume in the pipes is a good idea to adjust the mass flow? For the moment, the bigger part of the flow does not get in the pipes but passes on the upper and lower sides, this does not collapses the experimental data.

i apologise if my questions seem a little stupid but i'm a freshly new user and i'm learning at the moment... thanks again!

Strider

Re: problem with GGI interfaces
 

In rotating domain, total pressure is actually relative total pressure, while total pressure in stationary domain is absolute total pressure. Maybe this difference causes the total pressure drop across GGI interface.

Re: problem with GGI interfaces
 

Hi Strider,

Total Pressure is not a conserved quantity and is not energy. Momentum is conserved and either total or static enthalpy is conserved, but your total pressure can still change. Looking at the equation for total pressure in an incompressible fluid with constant Cp:

Pt = Ps + 1/2*Density*Velocity^2

Momentum is a conserved quantity and velocity is the momentum per unit mass. If you compute the average total pressure, you can see that it is locally varying by the velocity squared. If you mesh resolution changes, your total pressure will change because the relationship is non-linear and you are not including the entire velocity profile. You would have the same result if you sampled the mesh (or experimental data) at different resolutions and computed the total pressure.

What you are seeing is in essence the problem with mesh dependance. If your meshes were roughly the same size on either side of the GGI, the change in total pressure would be negligable. If the mesh resolution changed, but both meshes were sufficient for a mesh independant solution, you would also see very little change. However, since your mesh is not fine enough to capture all the details of the flow, you end up with some numerical diffusion and hence a loss in total pressure.

As I pointed out before, the same loss would occur if your mesh were fully connected but the resolution was reduced through a transition region. A GGI just makes it obvious because the change occurs at a 2D location.

Regards, Robin

Re: problem with GGI interfaces
 

Thank you Robin,

I understand the point. you would make a good teacher, if you are not already!

strider

Re: problem with GGI interfaces
 

Thanks for the compliment. I'm not a teacher, but I have had many similar discussions with CFX customers.