Discussion Forum

Hello,

I am modeling single phase, steady-state flow of incompressible, turbulent air through two pipes joined by a sharp 90 degree bend. The flow exhibits separation in two locations near the bend, and the flow should exhibit significant swirling about the pipe axis after flowing through the bend (I have attached an example of the model - flow enters the thinner pipe and exits the thicker pipe).

1) Which turbulence model would be most appropriate for this application? So far, I have tried Realizable k epsilon, k omega, and v2f models. I have stayed away from the standard k epsilon because of its alleged weakness in dealing with adverse pressure gradients, and I have not used the L-VEL and algebraic yPlus models because of their lower accuracies.

2) After obtaining a solution, I observe the streamlines and see that the flow loses swirl as it travels down the pipe after the bend. However, experiments reveal that there should be stronger swirl, and in theory, the lack of viscous effects outside the boundary layer should maintain the swirl far past the 90 degree bend. How can I obtain this swirl in my model?

3) I often see a very regular, periodic oscillation in the convergence plot, especially when I make the mesh finer. It is possible that this is because of time-dependent vortex behavior, but is there a way to deal with these oscillations in the convergence plot while continuing with stationary studies?

Any other tips would be highly appreciated.

Thank you!