Flow in a pipe: stability issues


I am trying to set-a flow in a 3D circular pipe based one the aneurysm bounceback case (especially for the boundary conditions).

I have a poiseuille flow at the inlet with a Reynolds number of around 400 and a pressure boundary at the outlet. I use bounceback boundary conditions for the wall.

The case runs but rapidly I get stability issues: non physical velocities appear along the wall and propagate in the flow until a got NANs !

I checked my stability parameters which seems right (U lb = 0.01). I run sensibilities for:

  • different lattices from around 1 to 40 millions (i. e. from 20 to 160 cells in a diameter)
  • different models: BGK or smagorinsky BGK model
  • different initializations: with default or poiseuille velocity initialization

Does everyone would have any idea of what is wrong, or what may set this right ? I am just out of ideas.

Thanks in advance,

I have no idea exactly but what I would do is to reduce the velocity at the inlet to smth like 0.0001 or less…

Anyway, if you post your code we can maybe be more helpful…

Good luck!


Hi sdea, and thank you for your quick answer.

I already tried to reduce the inlet velocity (to values around 0.0005) but without success. Nevertheless I think I figured out why my computation crashes: It comes from the voxalization of my STL pipe which creates something like that:

_ | _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ |
_ _ __ _ _ | _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ |
___________________ | _ |

So I have some bounceback nodes which are connected with almost only other bounceback node. This seems to be the problem since with a square pipe the computation runs without any trouble.

I will still post my sources so it may hep others.

Thank you,

Mmmm… I don’t know unfortunately I have very little experience in geometries from STL files. However, I work with porous media and I often encounter the situation you described. Sometimes, I have BB nodes which don’t see the fluid nodes at all!

In my case, the codes run smoothly…

Maybe somebody else can help you in this…


Hi sdea,

It seems you were right, I made tests with different lattices I do not have BB nodes in from of one another anymore but the computation still crashes… So I try your idea of reducing the velocity and the computation runs smoothly for Reynolds under 160 but crashes for higher Reynolds number…

I open a knew post for this specific problem “Lattice requirement for wall driven flow” with my code.

Thanks for your help,