[suktan]

Hi everyone,


I created a simple model for learning CFD. It is a fluid flow circulating around a circular loop in 2D (incompressible, laminar). It is created in midas-NFX CFD.

The flow is created by assigning velocity BC in an interior edge. The velocity is uniform across the loop. No-slip wall is assigned to inner and outer boundaries as depicted in the sketch.


Both steady and transient simulation converged. The velocity and pressure profile of steady simulation are shown in the graphs. They look like proper solution of the model.


Do you agree? Am I making a correct CFD exercise?


Thanks!

[sbaffini]

Not sure about how the code you are using handles your setup, but the proper way to simulate it would be to actually have no interior edge at all and use a volume source term for the two momentum equations, so that the combined effect of both is a resulting momentum source that for each location is tangential to the nearest wall

[suktan]

Thanks for your advice.

Originally Posted by sbaffini

use a volume source term

I am not sure whether I can define a volume source term in midas-NFX CFD. It is an FEM based code. Does it matter?

[sbaffini]

No, it shouldn't. Just the terminology, in this case, should then be different, it is a momentum source term.

If your circular pipe is centered in (x0,y0), for a given position (x,y), the angle that the relative position vector (x-x0,y-y0) makes with the x-aligned axis passing trough the center (x0,y0) is given by theta = atan2(y-y0,x-x0). Then, for a given magnitude m of the source term, you will need the following x and y components of momentum source terms for each position in the pipe:

mx = - m sin(atan2(y-y0,x-x0))
my = m cos(atan2(y-y0,x-x0))

given m, x0, y0. For m> 0 the flow will be couterclockwise. The higher m the higher the resulting velocity. A relation that links m to the mass flow rate in the pipe is, basically, a friction factor formula, that might or not exist for this very case, and will depend from the state of the flow, laminar or turbulent, and if turbulent from the very specific turbulence model you use. The simplest approach could be to go by trial and error, unless your needs require something more specific.