[Nick88]

Hi,

I have some troubles with a simulation with the flow-split boundary condition.

My geometry is similar to a T junction between two pipes, with an inlet and two outlets. The fluid is water, the Re number (and Mach) is quite low so the flow should be laminar, and I am using the segregated solver. The inlet is a velocity-inlet, while the two outlets are set to flow-split. I tried two RANS simulations, one steady and one unsteady.

In the steady simulation the residuals go down very well but, after a lot of iterations they start to increase and the solution diverges, because there is a reverse flow on the two outlets. Then I have tried with an unsteady simulation (because there is a vortex shedding near the junction of the two pipes), but I still get the reverse flow. I know that the flow-split doesn't support the reverse flow .

In both cases it starts from a cell near the wall of the pipes, where there are very thin prism cells (I need to solve a thin boundary layer for the transport of a passive scalar), and then it expands to the other cells of the outlet. The mesh is of good quality (4 million of cells, polyhedral + prism layer cells), and I have tried to use a low under-relaxation factor, but the reverse flow still occur. The outlets are far away from the junction (more than 20 diameters) so the flow should be straight near them; nothing changes if I use two longer pipes, or a coarser mesh near the outlets.

I don't know the pressure at the outlets, so I cannot use two pressure-outlets, and I need to prescribe the flow split between the two pipes. Now I am using one pressure-outlet and two velocity-inlet as boundary conditions, but I don't know if it is correct.
Can I use the velocity-inlet for an outlet?
Otherwise, what kind of boundary conditions could be suitable for this simulation?

[LuckyTran]

Originally Posted by Nick88

Hi,

I have some troubles with a simulation with the flow-split boundary condition.

My geometry is similar to a T junction between two pipes, with an inlet and two outlets. The fluid is water, the Re number (and Mach) is quite low so the flow should be laminar, and I am using the segregated solver. The inlet is a velocity-inlet, while the two outlets are set to flow-split. I tried two RANS simulations, one steady and one unsteady.

If flow is laminar, you should be using the laminar model instead of the full RANS equations. Try that and see if you still have convergence issues. If you still do, then I suggest looking at your mesh quality. Start with skewness angle < 85 degrees. If you have variable properties (polynomia, etc.), use constant property formulations, it improves the stability by a lot. You can add the variable property after you have a well converged simple problem.

Originally Posted by Nick88

Can I use the velocity-inlet for an outlet?
Otherwise, what kind of boundary conditions could be suitable for this simulation?

You can trick Star into using a velocity inlet or pressure inlet as an outlet, but this is not recommended. Flow split outlet seems to be the correct choice. Remember that the sum of flow-split values should equal 1, did you double-check those?

[Nick88]

Originally Posted by LuckyTran

If flow is laminar, you should be using the laminar model instead of the full RANS equations. Try that and see if you still have convergence issues. If you still do, then I suggest looking at your mesh quality. Start with skewness angle < 85 degrees. If you have variable properties (polynomia, etc.), use constant property formulations, it improves the stability by a lot. You can add the variable property after you have a well converged simple problem.

Sorry, I am already using the laminar model. The quality of the mesh is quite good, the maximus skewness angle is 78 degree. I have only constant properties.

Originally Posted by LuckyTran

You can trick Star into using a velocity inlet or pressure inlet as an outlet, but this is not recommended. Flow split outlet seems to be the correct choice. Remember that the sum of flow-split values should equal 1, did you double-check those?

The sum of flow-split values is 1 (0.2 and 0.8). I have tried to use the velocity-inlet as an outlet boundary condition, it seems to work. If I can use it, I will do so.