Discussion Forum

You should just set u_in and p_out. Do not overspecify the solution. So, do not specify u_out or p_in.

Thanks for your reply, James.

You are right about the boundary conditions at the inlet and outlet.

What I am asking is how to set the boundary conditions to the left bottom and the left top to make it numerically a closed loop.

Let me know whether I clearly state the problem.

Thank you!

You may need to tell us a little more about the problem and how you are modeling it. Is this a full 3D model ? Is there anything going on between the exit and entrance path of the secondary flow you are asking about at the bottom left and top left ? If there is no heat changes or pressure losses, and the flow is simply normal going in/out of these regions, you compute the flow exiting at the bottom left, then assume it is normal to the exit / entrance plane. Then you could use the coupling variables feature of COMSOL to create an entrance flow at the top left equal to the lower left. I suggest to start with simple 2D model first, then get more complex as you learn how to do it.

You may need to tell us a little more about the problem and how you are modeling it. Is this a full 3D model ? Is there anything going on between the exit and entrance path of the secondary flow you are asking about at the bottom left and top left ? If there is no heat changes or pressure losses, and the flow is simply normal going in/out of these regions, you compute the flow exiting at the bottom left, then assume it is normal to the exit / entrance plane. Then you could use the coupling variables feature of COMSOL to create an entrance flow at the top left equal to the lower left. I suggest to start with simple 2D model first, then get more complex as you learn how to do it.

Hi James, it is a 2D model. What I meant to ask is exact what you mentioned: coupling variables feature of COMSOL. Should I use 'Boundary similarity' to create an entrance flow at the top left equal to the lower left? But velocity is a vector, should I impose each component of velocity individually? And 'Boundary similarity' seems to only work when the values are the same, how can I make sure the direction of the velocity on these two boundaries are opposite?

Thanks soooooooooo much!

I think idmap would probably be what you need. It is also the simplest to use. You can experiment with all of them, and they all will probably work similarly. You can specify the normal velocity without having to specify each component. Are you worried about the velocity distribution within the boundary ? For example, if you coupling is accomplished with a pipe, it will have a distribution that changes along it's path. If you are now worried to that level of detail, you could just compute an average from the lower left using derived variables, then just set it equal to the normal velocity at the upper left boundary.

Will do!

Thank you!