[armanp]

Hi,
I am simulating a problem which had been solved in 2D and now trying to convert the code into 3D format to solve the 3D geometry. In this part of my udf (which I present below) I think I might have a problem because answers dont match the experiments like 2D.

face = Lookup_Thread(d, Cell_id);
begin_f_loop(f,face)
{
F_AREA(A,f,face);
Area = NV_MAG(A);
Cell_volt += F_UDSI(f,face,1)*Area;
Tot_Area += Area;
}
end_f_loop(f,face)
Cell_volt /= Tot_Area;

I'm new to udfs Do you guys know how I must change this part for getting results like 2D? Thanks in advance for your help

[`e`]

This code should work similarly in both 2-D and 3-D simulations. Have you double checked the dimensions, units and achieved mesh independence?

[armanp]

Thanks for your response. The model is mesh independent, I've checked that. But, about units in governing equations, I will check them again and get back to you.
Does this problem has anything with solver settings? I'm using PISO scheme. Also, with smaller Time-steps I have better convergence.

[`e`]

The problem may be with your solver settings, it doesn't appear to be the UDF at fault. Ensure you have also achieved time-step independence. Are there any 3-D simulations you can compare results with in the literature, perhaps the computational model is inadequate for describing the 3-D effects in this particular case?

[armanp]

I think the problem may relate to time-step size. Because, I can't achieve to an independent time-step model! When I raise the time step(up to 1 sec), my results diverge or get into a loop that it couldn't converge. Also, unfortunately there isn't a 3D model of my simulation that I could verify my results with that. There is just that 2D model!
I want to note that almost all of udfs in my 2d model was dimensionless, except a gradient function in which I added the z-direction component. Is it a good approach to just solve the model for a mesh that has been extruded in z for just one cell and then if the answers were good continue adding cells?

[`e`]

Yes, you could start with comparing your 3-D simulation with one cell depth versus your 2-D simulation. The results should match, bearing in mind the cell depth length.

What geometry are you simulating? You'll need to achieve time-step independence. Perhaps start with a simpler case (such as a geometry which has steady flow) to validate your model/workflow.

[armanp]

Thank you. I will try to simulate that and I'll get back to you.
My geometry is actually pretty simple. It is a box, divided into 4 parts. The Mesh around my walls and boundary conditions is intense.